bughz/anbox-modules
1
0
Fork 0
mirror of https://github.com/choff/anbox-modules.git synced 2026-10-17 16:55:48 +08:00
Code Issues Packages Projects Releases Wiki Activity

binder: don't relock after LRU_REMOVED_RETRY on Linux >= 6.13

Browse Source 
Since 6.13 the list_lru walker reacquires the lock itself after
LRU_REMOVED_RETRY. Taking it in the callback makes the next
lock_list_lru_of_memcg() spin forever, leaving kswapd at 100% CPU
in binder_shrink_scan().
This commit is contained in:
Anthony 2026-10-13 22:25:42 +02:00
parent 96d5d960fe
commit 0ee331db89
1 changed files with 9 additions and 1 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

10
binder/binder_alloc.c
View File

@ -1048,8 +1048,16 @@ enum lru_status binder_alloc_free_page(struct list_head *item,
trace_binder_unmap_kernel_end(alloc, index); trace_binder_unmap_kernel_end(alloc, index);
spin_lock(lock);
mutex_unlock(&alloc->mutex); mutex_unlock(&alloc->mutex);
#if (LINUX_VERSION_CODE < KERNEL_VERSION(6, 13, 0) && \
!defined(RHEL_10_1_BACKPORTS))
/* Before 6.13, LRU_REMOVED_RETRY means the lock was reacquired. */
spin_lock(lock);
#endif
/*
* Since 6.13 the walker relocks itself after LRU_REMOVED_RETRY;
* taking the lock here would make it spin forever (kswapd at 100%).
*/
return LRU_REMOVED_RETRY; return LRU_REMOVED_RETRY;
err_mmap_read_lock_failed: err_mmap_read_lock_failed: